Comparison of Breath-Hold vs. Free-Breathing Accelerated Cine with Deep Learning: Validation in Arrhythmia and Breath-Holding Patients

Motivation: Currently there is no consensus on whether breath-hold (BH) and free-breathing (FB) deep learning (DL) cine reconstruction techniques offer equivalent performance. Goal(s): To compare the clinical utility of accelerated DL image reconstruction in BH and FB cine, and identify the optimal DL-based cine for different clinical scenarios. Approach: Compare the performance of DL cine sequences across a general population and specific subgroups, including arrhythmia and dyspnea patients. Results: Both techniques ensure comparable quality and biventricular function with significantly reducing scan time. BH DL cine (at 10x acceleration) showed better performance in general group and arrhythmia subgroup; FB DL cine sequences benefits dyspnea patients. Impact: Deep learning cine reduce scan times while maintaining accurate biventricular parameters accuracy and comparable image quality, improving clinical work flow and patient comafort. Our research promotes their clinical use and provides guidance for selecting appropriate sequences in various clinical settings.